Analysis and modelling of tree succession on a recentrockslide deposit

Liesbet Van der Burght • Markus Stoffel •

Christof Bigler

Received: 29 August 2011 / Accepted: 7 November 2011 / Published online: 19 November 2011

� Springer Science+Business Media B.V. 2011

Abstract Dendroecological methods were com-

bined with vegetation and soil mapping to study

recolonisation of European larch (Larix decidua),

silver birch (Betula pendula) and Norway spruce

(Picea abies) on a recently formed rockslide cone

(deposit of 30 9 106 m3) in the Valais Alps (Switzer-

land). Tree density and tree height were predicted with

regression models that we derived using an informa-

tion-theoretic model selection approach. Results dem-

onstrate that the deposits of the 1991 rockslide have

been colonised relatively rapidly with larch (ecesis

time 2 years), birch (5 years) and spruce (2 years).

Most of the colonisation occurred 5–11 years follow-

ing the rockslide. Clast size was the primary factor

driving tree colonisation with the highest tree densities

found on plots with mainly smaller (\30 cm) clast

sizes. Tree height was affected by a combination of

different influences, with tree age and tree density

showing the most obvious effects. This study demon-

strates how dendroecological methods allow recon-

struction of spatio-temporal patterns of tree succession

on rockslides, which may ultimately facilitate a more

accurate dating of similar landforms of unknown age.

Keywords Rockslide � Dendroecology � Primary

succession � Ecesis � Regeneration � Betula pendula �Larix decidua � Picea abies

Introduction

Alpine regions are characterised by rugged topogra-

phy and geomorphic disturbances such as rockslides,

landslides, avalanches, glaciations or floods (Porter

and Orombelli 1980; Lewis and Smith 2004; Shroder

et al. 2012). Recently created surfaces are colonised by

vegetation passing through different successional

phases. These phases are initially controlled by factors

such as seed dispersal (e.g. distance to seed source,

wind direction, timing of seed release), weather

conditions, edaphic factors and erosion (Piotti et al.

2009), before other factors may become of importance

such as browsing by ungulates, competition and

pathogens (Cunningham et al. 2006).

In the past, several authors have used dendroeco-

logical methods to date surfaces created by geomor-

phic disturbances. However, a recurring problem

L. Van der Burght � M. Stoffel (&)

Laboratory of Dendrogeomorphology,

Institute of Geological Sciences, University of Berne,

Berne, Switzerland

e-mail: markus.stoffel@dendrolab.ch

M. Stoffel

Chair for Climatic Change and Climate Impacts,

Institute for Environmental Sciences,

University of Geneva, Geneva, Switzerland

C. Bigler

Forest Ecology, Institute of Terrestrial Ecosystems,

Department of Environmental Sciences, ETH Zurich,

Zurich, Switzerland

123

Plant Ecol (2012) 213:35–46

DOI 10.1007/s11258-011-0004-2

hindering accurate dating of geomorphic disturbances

has been the lack of data on ecesis, which is the time

interval between surface formation or stabilisation and

its colonisation by plants (Desloges and Ryder 1990;

McCarthy and Luckman 1993). Ecesis is species

specific and influenced by site climate (Pierson 2007).

Previous studies reported on ecesis values and colo-

nisation characteristics for deglaciated flood plains or

moraines (Bleuler 1986; Burga 1999; Garbarino et al.

2010; Masiokas et al. 2010; McCarthy and Luckman

1993), but only a very limited number of past studies

have focussed on the colonisation of surfaces created

by large-scale gravitational mass movements, which

were mostly limited to the re-establishment of vege-

tation on volcanic deposits (del Moral 1999; del Moral

and Ellis 2004; Lindig-Cisneros et al. 2006; Munoz-

Salinas et al. 2007; Pierson 2007). Data on tree

succession on rockslides or other kinds of talus

deposits is, in contrast, almost completely missing

(Wilson 1966; Walker et al. 1996).

In this study, we document primary succession of

trees and shrubs on the deposits formed by the

‘Grossgufer’ rockslide near Randa (Valais Alps,

Switzerland; Fig. 1). On 18 April 1991, a series of

rockslides deposited 20 9 106 m3 of gneissic rock and

debris in the valley floor, followed by another

10 9 106 m3 on 9 May 1991 (Schindler et al. 1993).

Although rockslide events are a recurrent process in

high mountain chains such as the Alps (Erismann and

Abele 2001), only few contemporary, well-docu-

mented examples exist where detailed data on the

geomorphic disturbance (Huggel et al. 2010) can be

linked with data on tree succession. Therefore, the

objective of this study was to gain insights into the

spatial and temporal colonisation of a rockslide by

European larch (Larix decidua), Norway spruce

(Picea abies) and silver birch (Betula pendula)

through (1) a detailed analysis of tree regeneration,

tree density and tree ages and (2) a quantification of

environmental factors controlling tree density and tree

height on the rockslide deposit.

Study site description

The ‘Grossgufer’ rockslide deposit is located in the

southern Swiss Alps (46�0604000N, 7�4605000E; Fig. 1)

and consists of gneissic and schist lithologies. The

rockslide cone has a surface of *68 ha and extends

from 1,300 to 1,900 m a.s.l. The site is characterised

by a continental climate with average annual rainfall

of *610 mm measured at the nearest climate station

located in Zermatt (46�0101000N, 7�4404700E, 1,618 m

a.s.l.; MeteoSwiss 2011). Winters are relatively short

compared to other sites at the same altitude with the

rockslide deposit being snow-covered from December

to March. Albedo effects of the deposit and the low

abundance of vegetation cover have an important

influence on local climatic conditions, which tend to

be harsher on the cone than elsewhere in the valley.

About half of the study site is dominated by large

blocks (C50% of clasts with Ø [ 1 m), which create

large, shady and relatively moist sites that are

preferred environments for several local plant species.

Amongst the tree and shrub species observed on the

deposit are larch, spruce, birch, green alder (Alnus

viridis), European red elder (Sambucus racemosa),

purple willow (Salix purpurea), goat willow (Salix

caprea), trembling aspen (Populus tremula) and

Fig. 1 The ‘Grossgufer’ rockslide deposit (Randa, Valais Alps,

Switzerland) was formed in April and May 1991 by a series of

massive rockfalls and rockslides depositing 30 9 106 m3 of

gneissic boulders, rocks and finer sediments

36 Plant Ecol (2012) 213:35–46

123

sycamore maple (Acer pseudoplatanus). The growing

season of trees in the region starts in late-May and ends

in early-October (Stoffel et al. 2005). Post-disturbance

colonisation by trees and shrubs occurred naturally

within the study area from nearby (\200–500 m) seed

sources. In the northern part of the rockslide deposit

(Fig. 1) seeding activity has influenced tree regener-

ation; we therefore excluded this sector and its

neighbourhood to avoid anthropogenic signals in the

analysis. In addition, broaching and anthropogenic

reshaping have deprived the eastern rim of the deposit

from its original character, which made sampling in

this part of the deposit redundant as well.

Fine-grained debris and silt are most frequent in the

northernmost and western segments of the study area;

large blocks dominate the south-eastern part (Fig. 1).

Between these two rather uniform areas, a transition

zone of medium and small blocks is observed. The

northern and southern borders of the deposit are funnel

shaped and dominated by fine-grained debris and silt.

In summary, along a north–south axis, a transition

exists from fine-grained debris and silt to large blocks

from the outer to the inner regions of the deposit.

Along an east–west axis, a less gradual transition from

large blocks to fine-grained debris and silt is observed.

Soils are scarce on the young deposit and generally

characterised by lime deficiency. However, a top layer

rich in leaf mould usually exists where vegetation is

present.

Methods

Mapping of substratum and vegetation

The area surveyed is located between 1,380 and

1,500 m a.s.l. (mean slope gradients: 14–32�) and

covers 12.3 ha, representing *18% of the rockslide

deposit. The surface was divided into homogeneous

plots of comparable size and plot boundaries defined

based on the grain size (GS) distribution of the

substratum and on tree density. Spatial data were

collected in the field using a Nikon 55DAS laser

distometer, compass and clinometer. For a total of 171

plots, 84 were composed of large blocks (GS0 C 50%

of clasts with Ø [ 1 m; Fig. 2), 40 of medium and

small blocks (GS1 C 50% of clasts with Ø 1–0.3 m),

and 47 of fine-grained debris and silt (GS2 C 50% of

clasts with Ø \ 0.3 m).

Collection of field data

Fieldwork was conducted in summer 2009 and spring

2010. For 93 of the 171 plots, the number of trees per

species was determined, and tree height and age were

assessed as described below. The broad range of plots

with varying tree composition and density provided

sufficient variation to individually compare these with

the remaining 78 plots. The relative cover of herbs and

shrubs was visually estimated at the plot level.

Tree ages were determined by either (1) counting

tree rings at the height of the root collar after

destructive sampling of trees or by (2) counting bud-

scale scars on the stem surface (Hoffmann and

Schweingruber 2002). The height of larch trees was

measured and approximately three trees per plot (i.e.

269 trees within the 93 measured plots) felled. The

bud-scale scar method was applied to all larches (i.e.

1,792 trees), except for densely populated plots where

trees were visually assigned to several classes of tree

height. In addition to the age determination of every

larch by counting its bud-scale scars, 10% of the trees

of the respective height class were felled and growth

rings were counted to validate the accuracy of the bud-

scale scar approach. For birch, the height of each

individual was measured, unless (as in the case of

larch) for plots with a high number of trees. A total of

39 birch trees of different height were felled in

randomly selected plots to get an estimate of their age.

For spruce, tree height was measured as well and

approximately 10% of all specimens were felled per

plot (i.e. 45 trees) to determine their age.

Field data were compiled in a database containing

information: (1) at the plot level (substratum, total

number of trees, number of trees per species, area and

tree density); and (2) at the tree level (species, height,

bud-scale scar age and tree-ring age). At the plot level

topographic variables retrieved from ArcGIS v10,

including altitude, slope, aspect and proximity to the

nearest forest edge, were added (Table 1).

Modelling tree density and height

To model the number of trees per plot, a generalized

linear model assuming a negative binomial distribu-

tion (Crawley 2007) was fitted to the birch, larch and

spruce populations. The negative binomial distribu-

tion allows to model count data with overdispersion,

i.e. the variance is greater than the mean (Venables

Plant Ecol (2012) 213:35–46 37

123

and Ripley 2002). The probability density function of

the negative binomial distribution is defined for counts

y as:

f y; h; lð Þ ¼ C hþ yð ÞC hð Þ � y!

� lyhh

lþ hð Þhþyð1Þ

where C is the gamma distribution and h is the

dispersion parameter. The mean l is modelled as:

log lð Þ ¼ log Að Þ þ Xb ð2Þ

where log(A) is a vector with the log-transformed plot

areas, X is a matrix that contains the predictor

variables and b is a vector with the regression

coefficients. Substratum, altitude, slope, aspect and

proximity to the nearest forest edge were set as

predictor variables to model the number of trees per

plot. For the model selection, we used an information-

Fig. 2 Grain-size (GS) distribution on the study site located in

the lower reaches of the rockslide deposit: GS0 is composed of

large blocks (C50% of clasts with Ø [ 1 m), GS1 has

predominantly medium and small blocks (C50% of clasts with

Ø 1–0.3 m) and GS2 is characterised by fine-grained debris and

silt (C50% of clasts with Ø \ 0.3 m)

Table 1 Description of variables at the tree and plot level

Variables Level Description

Species Tree Larix decidua/Betula pendula/Picea abies/Alnus viridis/Sambucusracemosa/Salix purpurea/Salix caprea/Populus tremula/Acer pseudoplatanus

Tree height Tree Ranging from 0.03 to 7 m

Growth-ring age Tree Ranging from 1 to 17 years

Bud-scale scar age Tree Ranging from 1 to 17 years

Substratum (grain size) Plot Large blocks (GS 0): C50% of clasts with Ø [ 1 m

Medium and small blocks (GS 1): C50% of clasts with Ø 1–0.3 m

Fine-grained debris and silt (GS 2): C50% of clasts with Ø \ 0.3 m

Number of trees per species Plot Ranging from 0 to 1,526 trees

Total number of trees Plot Ranging from 0 to 2,877 trees

Tree density Plot Ranging from 0 to 6,691 trees/ha

Area Plot Ranging from 57 to 2,356 m2

Altitude Plot Ranging from 1382.5 to 1488.9 m a.s.l.

Slope Plot Ranging from 14.3 to 32.4�Distance to nearest forest edge Plot Ranging from 217.7 to 512.0 m

Aspect Plot Northeast-facing slopes (NE): 22–66�East-facing slopes (E): 67–111�Southeast-facing slopes (SE): 112–157�

Growth-ring ages were assessed for larch, birch and spruce as described in Collection of field data. Bud-scale scar ages were assessed

for larch as described in Collection of field data

38 Plant Ecol (2012) 213:35–46

123

theoretic approach based on the AICc (corrected

Akaike Information Criterion) (Burnham and Ander-

son 2002; Stauffer 2008):

AICc ¼ �2� log likelihoodð Þ þ 2� p� n

n� p� 1

ð3Þ

where p is the number of parameters used in the model

and n is the sample size. We selected a priori nine

different models that contained between one and five

predictor variables (Table 2); the model with the

lowest AICc amongst these candidate models is the

best-fitting model. Unlike other model selection

approaches such as step-wise model selection, the

information-theoretic approach does not emphasize

significance testing (Stauffer 2008). Akaike weights wi

were then calculated, which are the probability that

model i is the best-fitting model amongst the nine

models (Burnham and Anderson 2002):

wi ¼exp � 1

2Di

� �

P9j¼1 exp � 1

2Dj

� � ð4Þ

where Di designates the difference in AICc between

model i and the model with the lowest AICc.

To model tree height, log-linear models were fitted

to the birch, larch and spruce populations:

log heightð Þ ¼ Xb ð5Þ

where height is tree height in 2009/2010, X is a matrix

with the predictor variables and b is a vector that

contains the regression coefficients. Age, substratum,

tree density, altitude, slope and aspect were used as

predictor variables for explaining tree height

(Table 3). Similar to the model selection approach

for modelling tree density, we selected a priori nine

different models (Table 3) that we compared based on

the AICc (Eq. 3) and Akaike weights (Eq. 4). Since,

tree age is expected to be an important variable in

predicting tree height, we included the variable age

(estimated with tree rings and bud-scale scars for larch

and with tree rings for birch and spruce) in any of the

models. For birch and spruce, we calculated fixed-

effects models (Eq. 5). For larch, we calculated

mixed-effects models (Pinheiro and Bates 2000) since

several trees per plot were sampled, i.e. the height data

were assumed to be dependent within each plot. Thus,

to account for plot-to-plot variability, random effects

were added for both the intercept and the variable age Ta

ble

2M

od

else

lect

ion

of

neg

ativ

eb

ino

mia

lm

od

els

for

pre

dic

tin

gth

en

um

ber

of

tree

sp

erp

lot

Mo

del

nu

mb

er

Var

iab

les

Bir

ch(n

=1

71

)L

arch

(n=

17

1)

Sp

ruce

(n=

17

1)

Su

bst

ratu

mA

ltit

ud

eS

lop

eA

spec

tP

rox

.to

fore

st

AIC

cD

AIC

cA

kai

ke

wei

gh

t

(%)

AIC

cD

AIC

cA

kai

ke

wei

gh

t

(%)

AIC

cD

AIC

cA

kai

ke

wei

gh

t

(%)

1x

18

61

.10

.07

11

82

4.7

0.7

35

10

05

.10

.07

9

2x

19

30

.06

8.9

01

91

1.1

87

.10

10

74

.86

9.7

0

3x

19

17

.85

6.6

01

89

0.9

66

.90

10

62

.05

6.9

0

4x

19

12

.15

1.0

01

88

0.5

56

.50

10

61

.05

5.9

0

5x

19

27

.76

6.5

01

90

4.6

80

.60

10

72

.36

7.2

0

6x

xx

19

04

.64

3.4

01

86

9.9

45

.90

10

44

.83

9.7

0

7x

xx

x1

86

3.9

2.8

18

18

24

.00

.04

91

00

8.4

3.3

15

8x

xx

x1

90

6.7

45

.60

18

70

.14

6.1

01

04

5.4

40

.30

9x

xx

xx

18

64

.93

.71

11

82

6.2

2.2

16

10

10

.55

.45

Sh

ow

nar

eth

ep

red

icto

rv

aria

ble

sin

clu

ded

inth

em

od

els

for

bir

ch,

larc

han

dsp

ruce

.T

he

nu

mb

ero

fo

bse

rvat

ion

s(n

)is

ind

icat

edin

par

enth

eses

.A

ICc

isth

eco

rrec

ted

AIC

(Ak

aik

eIn

form

atio

nC

rite

rio

n),

DA

ICc

isth

ed

iffe

ren

ceo

fth

eA

ICc

toth

em

od

elw

ith

the

low

est

AIC

can

dth

eA

kai

ke

wei

gh

tis

the

pro

bab

ilit

yo

fa

mo

del

bei

ng

the

bes

t-fi

ttin

g

mo

del

wit

hin

the

list

of

mo

del

s.A

llm

od

els

wer

efi

tted

usi

ng

max

imu

mli

kel

iho

od

(ML

)

Plant Ecol (2012) 213:35–46 39

123

(Eq. 5), which were grouped by plot (Pinheiro and

Bates 2000).

Statistical analyses and modelling of the data was

conducted with the R software (R Development Core

Team, 2010; Version 2.11.1). To fit negative binomial

models to the tree density data, we used the function

‘glm.nb’ in the R package ‘MASS’ (Venables and

Ripley 2002). To predict tree height based on log-

linear models, we used the ‘gls’ function from the R

package ‘nlme’ for fitting fixed-effects models to the

birch and spruce data and the ‘lme’ function for fitting

mixed-effects models to the larch data (Pinheiro and

Bates 2000).

Results

Tree abundance and density

Larch (*45% of the entire population) and birch

(*46%) were clearly the dominant tree species on the

rockslide deposit and within the different substratum

types (Fig. 3a, b). Differences between their respec-

tive abundances seemed to diminish with grain size:

from*10% for the large blocks (GS0) to*1% for the

medium and small blocks (GS1) with birch being more

abundant in both substratum classes. The smallest

difference in abundance of *0.5% was observed for

fine-grained debris and silt (GS2) with larch being

more abundant. Spruce (*6% of the entire popula-

tion; Fig. 3c) clearly lags behind larch and birch, as do

the other species (*3%). An average density of 4,062

trees ha-1 was registered with the highest densities

estimated on plots of fine-grained debris and silt

(Fig. 2). Densities clearly increased with decreasing

grain size and were highest for the three most abundant

species in the northern and western areas of the deposit

(Fig. 3).

Primary succession

The earliest regeneration occurred 2 years after the

rockslide in 1993 for larch growing on GS1. The oldest

larch growing on GS0 substratum is from 1994 and on

GS2 from 1995. An ecesis value of 5 years (1996) was

obtained for birch, ecesis values of 2 years were found

for spruce.

The development of the larch population was

reconstructed by mapping the number of seedlingsTa

ble

3M

od

else

lect

ion

of

log

-lin

ear

mo

del

sfo

rp

red

icti

ng

tree

hei

gh

t

Mo

del

nu

mb

erV

aria

ble

sB

irch

(n=

39

)L

arch

(n=

1,7

92

)S

pru

ce(n

=4

5)

Ag

eS

ub

stra

tum

Tre

e

den

sity

Alt

itu

de

Slo

pe

Asp

ect

AIC

cD

AIC

cA

kai

ke

wei

ght

(%)

AIC

cD

AIC

cA

kai

ke

wei

gh

t(%

)A

ICc

DA

ICc

Ak

aike

wei

gh

t(%

)

1x

x8

1.3

26

.40

35

20.1

13

.10

71

.60

.05

5

2x

58

.63

.75

35

24.3

16

.70

76

.44

.85

3x

x5

8.0

3.1

73

52

0.9

13

.30

72

.61

.03

4

4x

xx

58

.23

.46

35

24.6

17

.00

76

.95

.24

5x

xX

X6

6.2

11

.40

35

17.0

9.4

18

1.0

9.4

1

6x

xx

XX

54

.80

.03

33

51

2.9

5.3

48

4.1

12

.50

7x

xX

X6

5.8

11

.00

35

13.0

5.4

47

8.9

7.3

1

8x

xx

XX

56

.51

.61

53

50

7.6

0.0

57

81

.91

0.2

0

9x

xx

xX

X5

4.8

0.0

33

35

08.6

1.0

34

87

.21

5.6

0

Sh

ow

nar

eth

ep

red

icto

rv

aria

ble

sin

clu

ded

inth

em

od

els

for

bir

ch,

larc

han

dsp

ruce

.T

he

nu

mb

ero

fo

bse

rvat

ion

s(n

)is

indic

ated

inp

aren

thes

es.

AIC

cis

the

corr

ecte

dA

IC(A

kai

ke

Info

rmat

ion

Cri

teri

on

),D

AIC

cis

the

dif

fere

nce

of

the

AIC

cto

the

model

wit

hth

elo

wes

tA

ICc

and

the

Akai

ke

wei

ght

isth

epro

bab

ilit

yo

fa

model

bei

ng

the

bes

t-fi

ttin

gm

odel

wit

hin

the

list

of

mod

els.

Mo

del

sfi

tted

for

bir

chan

dsp

ruce

are

fix

ed-e

ffec

tsm

od

els,

mod

els

fitt

edfo

rla

rch

are

mix

ed-e

ffec

tsm

od

els

incl

ud

ing

inte

rcep

tan

dag

eas

random

effe

cts.

All

model

sw

ere

fitt

ed

usi

ng

max

imu

mli

kel

ihoo

d(M

L)

40 Plant Ecol (2012) 213:35–46

123

that germinated in a distinct period as a percentage of

the current population (Fig. 4). In 1993 and 1994, the

surveyed area remained largely void of trees apart

from some plots in the north-western segment of the

study area, where up to 14% of the current larch

population had germinated (Fig. 4a). Tree expansion

was much more intense between 1995 and 1998, when

up to 100% of the current population germinated

within some plots. In contrast, a lack of larch

germination is still clearly visible in the plots with

GS0 between 1995 and 1998 (Fig. 4b). For the period

1999–2002, high colonisation rates were observed in

the central part of the study area and plots of any grain

size (including GS0) were colonised (Fig. 4c). Colo-

nisation rates diminished again from 2007 to 2010; a

maximum of 22% of the current population estab-

lished in some plots (Fig. 4e).

Modelling tree density

The model with substratum as the only predictor

variable was by far the best-fitting model for predict-

ing tree density for both birch (Akaike weight of 71%)

and spruce (Akaike weight of 79%) (model 1;

Table 2). For both species, the number of trees per

plot increased significantly with decreasing grain size

(model 1; Table 4). Other models for birch and spruce

with Akaike weights between 5 and 18% additionally

included altitude, slope, aspect and proximity to the

nearest forest edge (models 7 and 9; Tables 2, 4). For

larch there were relatively small differences in terms

of Akaike weights between models 7 (49%) and 1

(35%; Table 2). Similar to larch and birch, both

models showed the same effect of substratum on tree

density. For model 7, tree densities tended to increase

with altitude (Table 4). Similar to larch and birch, tree

densities tended to be higher on northeast-facing

slopes compared to east-facing slopes.

Modelling tree height

Tree height of birch was equally well fit by models 6

and 9 (Akaike weight 33%) with model 6 including

age, substratum, altitude, slope and aspect, and model

9 additionally including tree density (Table 3). Tree

height tended to decrease from GS0 to GS1 to GS2,

whereas tree height strongly increased with age and

altitude (Table 5). Slope and aspect had only marginal

effects on tree height. Increasing tree density tended to

decrease tree height (model 9; Table 5).

Fig. 3 Stand densities (in

103 trees ha-1) on the

rockslide deposit as

observed in 2010 for a larch,

b birch and c spruce

Plant Ecol (2012) 213:35–46 41

123

Fig. 4 Spatial distribution

of freshly sprouted larch

seedlings over given time

intervals. Changes as

compared to the current

(2010) larch population are

expressed as a percentage:

a 1993–1994, b 1995–1998,

c 1999–2002, d 2003–2006

and e 2007–2010

42 Plant Ecol (2012) 213:35–46

123

For larch, models 8 and 9 with Akaike weights of

57 and 34%, respectively, were the best-fitting models

with model 9 including all predictor variables and

model 8 leaving out only substratum (Table 3). Older

trees and trees growing higher up on the deposit and on

northeast-facing slopes were growing taller; there

were also some positive but marginal effects from

substratum and non-significant effects from slope on

tree height (Table 5). Tree density had a negative

effect on tree height. There was a relatively high plot-

to-plot variability for the relationship between age

and tree height as indicated by the random effects

(Table 5).

For spruce, models 1 and 3 were the best-fitting

models with Akaike weights of 55 and 34%, respec-

tively (Table 3). Older trees tended to be taller and tree

density had a weak negative effect on tree height

(Table 5).

Discussion and conclusions

This study reports on the analysis and modelling of

tree succession on a young (20-year old) *68 ha

rockslide surface. Our data illustrate that colonisation

of the rockslide deposit with trees occurred almost

immediately after the event: *1% of the current tree

population was present on the deposit by 3 years after

the event, *38% of the population colonised the

surface within the first 7 years and *83% of the

current tree population germinated within 11 years.

The results represent one of the first datasets

analysing tree succession on a rockslide deposit, i.e.

a surface which is almost bare of any soil and

composed of blocky, sandy and silty debris of gneissic

origin. A comparison with existing studies on primary

succession, which have been conducted in different

contexts and on different deposits, is therefore chal-

lenging. To a certain degree the ecesis data from our

study can be compared with the results of Pierson

(2007), who analysed tree colonisation on lahar

deposits, where ecesis values ranged from 1 to

14 years. In his study, seed sources were at compa-

rable distances from the study site. However, the

deposits on which ecesis was analysed were of

different nature (more acid petrology and smaller

granulometry) and therefore resulted in much lower

albedo. The amount of precipitation differs as well

between the two sites, with the Mount St. HelensTa

ble

4M

od

eld

escr

ipti

on

so

fth

en

egat

ive

bin

om

ial

mo

del

sfo

rp

red

icti

ng

the

nu

mb

ero

ftr

ees

per

plo

t

Sp

ecie

sM

od

eln

um

ber

Inte

rcep

tS

ub

stra

tum

(GS

1)

Su

bst

ratu

m(G

S2

)A

ltit

ud

eS

lop

eA

spec

t(N

E)

Asp

ect

(SE

)D

isp

ersi

on

par

amet

er

Bir

ch1

-2

.82

7±

0.1

26

(\0

.00

1)

1.5

28

±0

.22

0

(\0

.00

1)

1.8

10

±0

.20

8

(\0

.00

1)

0.7

74

±0

.07

7

Bir

ch7

-1

2.0

94

±6

.08

2

(0.0

47

)

1.4

09

±0

.25

6

(\0

.00

1)

1.6

00

±0

.23

1

(\0

.00

1)

0.0

06

±0

.00

4

(0.1

37

)

0.0

13

±0

.02

6

(0.6

12

)

0.7

51

±0

.29

0

(0.0

10

)

0.0

64

7±

0.2

58

(0.8

02

)

0.7

97

±0

.08

0

Lar

ch1

-3

.07

6±

0.1

25

(\0

.00

1)

1.8

07

±0

.21

9

(\0

.00

1)

2.0

73

±0

.20

7

(\0

.00

1)

0.7

85

±0

.07

9

Lar

ch7

-1

1.0

10

±5

.99

2

(0.0

66

)

1.4

00

±0

.25

2

(\0

.00

1)

1.6

70

±0

.22

8

(\0

.00

1)

0.0

06

±0

.00

4

(0.1

59

)

-0

.00

8±

0.0

26

(0.7

54

)

0.7

31

±0

.28

5

(0.0

10

)

-0

.21

7±

0.2

54

(0.3

93

)

0.8

23

±0

.08

3

Sp

ruce

1-

5.7

82

±0

.19

2

(\0

.00

1)

2.0

82

±0

.32

0

(\0

.00

1)

3.0

00

±0

.30

2

(\0

.00

1)

0.3

97

±0

.05

3

Sp

ruce

7-

16

.61

9±

8.7

06

(0.0

56

)

1.7

45

±0

.37

1

(\0

.00

1)

2.5

20

±0

.33

1

(\0

.00

1)

0.0

08

±0

.00

6

(0.1

67

)

-0

.03

6±

0.0

38

(0.3

48

)

0.6

34

±0

.40

9

(0.1

21

)

-0

.20

8±

0.3

74

(0.5

78

)

0.4

17

±0

.05

7

Fo

rea

chsp

ecie

s,th

etw

ob

est-

fitt

ing

mo

del

s(c

f.T

able

2)

are

sho

wn

Sh

ow

nar

ees

tim

ates

±st

and

ard

erro

rs(P

val

ues

)fo

rth

ev

aria

ble

s;fo

rth

ed

isp

ersi

on

par

amet

er,

esti

mat

es±

stan

dar

der

rors

are

sho

wn

.F

or

sub

stra

tum

and

asp

ect,

resp

ecti

vel

y,

larg

eb

lock

s(G

S0

)an

dea

st-f

acin

gsl

op

es(E

)ar

ese

tas

refe

ren

cele

vel

s.A

llm

od

els

wer

efi

tted

usi

ng

max

imu

mli

kel

iho

od

(ML

)

Plant Ecol (2012) 213:35–46 43

123

region (Pierson 2007) receiving almost twice as much

precipitation as compared to Randa. Much greater

differences exist between our data from the rockslide

deposit and ecesis in glacier forefields, where coloni-

sation has been studied rather extensively (Bleuler

1986; Burga 1999; Garbarino et al. 2010; McCarthy

and Luckman 1993). A comparison of ecesis data from

glacier forefields with our data set shows 3–17 times

longer lag times, probably reflecting the hampering of

tree colonisation in glacier forefields by cool glacial

winds, meandering meltwater streams and the scarcity

of seed sources.

The large amount of dust produced by the rockslide

probably accelerated seedling establishment. Although,

the rockslide deposit was not covered with soil sensu

stricto, the 400,000–1,000,000 m3 (Schindler et al. 1993)

of dust arising from the event provided a fertile fine

substrate for seedling establishment. The rockslide

deposit was colonised primarily (*91%) by the pioneer

species larch and birch reflecting their ability to colonise

new surfaces (Ellenberg 1988). The number of spruces

identified on the deposit (6%) was, in contrast, more

unusual, since this species tends to occur during

secondary successional phases (Ellenberg 1988).

Irrespective of the species, colonisation of the

deposit was predominantly determined by the vari-

ability of the substratum (Table 2). Tree density of

birch, larch and spruce significantly increased from

plots with large blocks to plots with fine-grained

debris and silt (Table 4). For larch, additional envi-

ronmental influences including altitude, slope and

aspect affected tree density. However, only northeast-

facing plots showed a significant increase in tree

density. For all three species, this effect may not

simply be attributed to the apparent correlation of

substratum and aspect (Fig. 2), because any model

including substratum as predictor variable achieved a

relatively high Akaike weight, which was not the case

for models including aspect but not substratum

(Table 2). The slight increase of tree density with

altitude (model 7, Table 4) should not be attributed to

a classical altitudinal effect (i.e. decreasing tempera-

ture or increasing precipitation, respectively, with

increasing altitude) as the difference in altitude is not

large enough between the lower and the upper plots,

but most likely reflects changes in soil moisture and

nutrient availability or insolation. For all three tree

species, distance to the closest forest edge was not of

importance for the study site in Randa, probably as aTa

ble

5M

od

eld

escr

ipti

on

so

fth

elo

g-l

inea

rm

od

els

for

pre

dic

tin

gtr

eeh

eig

ht

Spec

ies

Model

num

ber

Fix

edef

fect

sR

andom

effe

cts

Inte

rcep

tA

ge

Subst

ratu

m

(GS

1)

Subst

ratu

m

(GS

2)

Tre

eden

sity

Alt

itude

Slo

pe

Asp

ect

(NE

)A

spec

t(S

E)

Inte

rcep

tA

ge

Bir

ch6

-19.1

29

±4.9

79

(0.0

01)

0.0

68

±0.0

28

(0.0

23)

-0.1

24

±0.3

58

(0.7

32)

-1.1

26

±0.3

23

(0.0

02)

0.0

13

±0.0

04

(0.0

01)

0.0

21

±0.0

34

(0.5

35)

0.1

43

±0.2

30

(0.5

39)

-0.4

62

±0.4

09

(0.2

67)

Bir

ch9

-18.4

52

±4.8

47

(0.0

01)

0.0

87

±0.0

30

(0.0

07)

-0.0

42

±0.3

50

(0.9

05)

-0.8

02

±0.3

66

(0.0

36)

-0.4

72

±0.2

75

(0.1

00)

0.0

13

±0.0

03

(0.0

01)

0.0

18

±0.0

33

(0.5

90)

0.2

74

±0.2

36

(0.2

56)

-0.4

40

±0.4

00

(0.2

76)

Lar

ch8

-13.8

04

±4.2

01

(0.0

01)

0.2

42

±0.0

11

(\0.0

01)

-0.3

37

±0.1

47

(0.0

24)

0.0

08

±0.0

03

(0.0

08)

0.0

16

±0.0

15

(0.3

00)

0.7

62

±0.2

48

(0.0

03)

0.1

51

±0.1

32

(0.2

58)

0.8

20

0.0

64

Lar

ch9

-14.3

33

±4.2

63

(0.0

01)

0.2

42

±0.0

11

5(\

0.0

01)

0.2

80

±0.1

61

(0.0

86)

0.0

94

±0.1

74

(0.5

89)

-0.4

46

±0.1

81

(0.0

16)

0.0

08

±0.0

03

(0.0

06)

0.0

13

±0.0

16

(0.3

93)

0.7

43

±0.2

51

(0.0

04)

0.2

61

±0.1

49

(0.0

84)

0.7

95

0.0

62

Spru

ce1

-1.6

53

±0.4

12

(\0.0

01)

0.0

56

±0.0

32

(0.0

83)

Spru

ce3

-1.5

59

±0.4

18

(0.0

01)

0.0

60

±0.0

32

(0.0

66)

-0.2

23

±0.1

91

(0.2

49)

For

each

spec

ies,

the

two

bes

t-fi

ttin

gm

odel

s(c

f.T

able

3)

are

show

n

For

the

fixed

effe

cts,

esti

mat

es±

stan

dar

der

rors

(Pval

ues

)ar

esh

ow

n;

for

the

random

effe

cts,

stan

dar

ddev

iati

ons

are

show

n.F

or

subst

ratu

man

das

pec

t,re

spec

tivel

y,

larg

eblo

cks

(GS

0)

and

east

-fac

ing

slopes

(E)

are

set

asre

fere

nce

level

s.T

he

fixed

-eff

ects

model

sfo

rbir

chan

dsp

ruce

wer

efi

tted

usi

ng

max

imum

likel

ihood

(ML

),th

em

ixed

-eff

ects

model

sfo

rla

rch

wer

efi

tted

usi

ng

rest

rict

edm

axim

um

likel

ihood

(RE

ML

)

44 Plant Ecol (2012) 213:35–46

123

result of the light-weight seeds of the three anemoch-

orous tree species and because the closest forest edges

were found within 200–500 m from the plots. Seed

dispersal e.g. for Norway spruce has been reported to

occur over several 100 m from the seed source (Piotti

et al. 2009).

Important influences on tree height of birch, larch

and spruce were tree age and tree density (Table 3). As

expected, older trees were taller, whereas tree height

decreased with increasing tree density (Table 5). Tree

competition seemed to be of importance already less

than 20 years following the rockslide, and more so for

the shade-intolerant birch and larch than for the

intermediate shade-tolerant spruce. For the two pioneer

species birch and larch, substratum, altitude, slope, and

aspect turned out to be further variables that partially

determine tree height. Tree height of birch was reduced

on plots with fine-grained debris and silt, i.e. birches in

plots with large blocks tended to be larger maybe as a

result of the protective effect of the rocks against snow

movements. Similar to the effect of altitude on tree

density, the significant effect of altitude on tree height

might be due to some unmeasured variables that are

confounded with altitude such as soil moisture. For

larch, northeast-facing slopes seem to be more benefi-

cial for height growth than east-facing slopes (Table 5).

Because larch established earlier on northeast-facing

plots than on east-facing plots (results not shown), these

trees might have had a competitive advantage com-

pared to trees on adjacent plots.

By providing detailed recolonization data for larch

and by identifying important environmental influences

that affect tree density and the development of tree

height for larch, birch and spruce, we increase the

general understanding of early tree succession pro-

cesses on rockslide deposits. Future studies on ecesis

and landform dating may benefit from our findings,

which indicate what processes drive ecesis times and

early successional phases in similar settings. Our

results demonstrate how dendroecological methods

allow reconstructing spatio-temporal patterns of tree

succession on a rockslide, which may ultimately

facilitate a more accurate dating of rockslide deposits

and similar landforms of unknown age.

Acknowledgments This study has benefited from field

assistance of Oliver Fux. We acknowledge the local forester

Leo Jorger for the sampling permit. We would like to thank

Associate Editor Chris Lusk and two reviewers for providing

useful comments on the manuscript. We dedicate this study to

our landlady Lina Summermatter (1926–2010)—in loving

memory, may you rest in peace.

References

Bleuler RM (1986) Jahrringanalysen von Larchen in Glet-

schervorfeldern. Diploma thesis, Geographical Institute,

University of Zurich

Burga CA (1999) Vegetation development on the glacial fore-

field morteratsch (Switzerland). Appl Veg Sci 2:17–24

Burnham KP, Anderson DR (2002) Model selection and mul-

timodel inference: a practical-theoretic approach. Springer,

New York

Crawley MJ (2007) The R book. Wiley, Chichester

Cunningham CA, Zimmerman NE, Stoeckli V, Bugmann H

(2006) Growth response of Norway spruce saplings to

artificial browsing, black snow mold, and ground vegeta-

tion in two forest gaps in the Swiss Alps. Can J For Res

36:2782–2796

del Moral R (1999) Plant succession on pummice at Mount St

Helens, Washington. Am Midl Nat 141:101–104

del Moral R, Ellis EE (2004) Gradients in compositional vari-

ation on lahars, Mount St. Helens, Washington, USA. Plant

Ecol 175:273–286

Desloges JR, Ryder JM (1990) Neoglacial history of the Coast

Mountains near Bella Coola, British Columbia. Can J Earth

Sci 27:281–290

Ellenberg H (1988) Vegetation ecology of central Europe.

Cambridge University Press, New York

Erismann TH, Abele G (2001) Dynamics of rockslides and

rockfalls. Springer, Berlin

Garbarino M, Lingua E, Nagel TA, Godone D, Motta R (2010)

Patterns of larch establishment following deglaciation of

Ventina glacier, Central Italian Alps. For Ecol Manag

259:583–590

Hoffmann S, Schweingruber FH (2002) Light shortage as a

modifying factor for growth dynamics and wood anatomy

in young deciduous trees. IAWA J 23:121–141

Huggel C, Salzmann N, Allen S, Caplan-Auerbach J, Fischer L,

Haeberli W, Larsen C, Schneider D, Wessels R (2010)

Recent and future warm extreme events and high-mountain

slope failures. Philos Trans Roy Soc A 368:2435–2459

Lewis D, Smith DJ (2004) Little Ice Age glacial activity in

Strathcona Provincial Park, Vancouver Island, British

Columbia, Canada. Can J Earth Sci 41:285–297

Lindig-Cisneros R, Galindo-Vallejo S, Lara-Cabrera S (2006)

Vegetation on tephra deposits 50 years after the end of the

eruption of the Paricutın Volcano, Mexico. Southwest Nat

51:455–461

Masiokas MH, Luckman BH, Villalba R, Ripalta A, Rabassa J

(2010) Little Ice Age fluctuations of Glaciar Rio Manso in

the north Patagonian Andes of Argentina. Quat Res

73:96–106

McCarthy DP, Luckman BH (1993) Estimating ecesis for tree-

ring dating of moraines - a comparative study from the

Canadian Cordillera. Arct Alp Res 25:63–68

Munoz-Salinas E, Manea VC, Palacios D, Castillo-Rodriguez M

(2007) Estimation of lahar flow velocity on Popocatepetl

volcano (Mexico). Geomorphology 92:91–99

Plant Ecol (2012) 213:35–46 45

123

Pierson TC (2007) Dating young geomorphic surfaces using age

of colonizing Douglas fir in southwestern Washington and

northwestern Oregon, USA. Earth Surf Process Landf

32:811–831

Pinheiro JC, Bates DM (2000) Mixed-effects models in S and

S-Plus. Springer, New York

Piotti A, Leonardi S, Piovani P, Scalfi M, Menozzi P (2009)

Spruce colonization at treeline: where do those seeds come

from? Heredity 103:136–145

Porter SC, Orombelli G (1980) Catastrophic rockfall of Sep-

tember 12, 1717 on the Italian flanc of the Mont Blanc

massif. Zeitschrift fur Geomorphologie 24:200–219

Schindler C, Cuenod Y, Eisenlohr T, Joris C-L (1993) Die Er-

eignisse vom 18. April und 9. Mai 1991 bei Randa (VS) -

ein atypischer Bergsturz in Raten. Eclogae Geologicae

Helveticae 86:643–665

Shroder J Jr, Marston R, Stoffel M (2012) Treatise on geo-

morphology: mountain and hillslope geomorphology.

Academic Press, San Diego, vol. 7, in press

Stauffer HB (2008) Contemporary Bayesian and Frequentist

statistical research methods for natural resource scientists.

Wiley, Chichester

Stoffel M, Lievre I, Monbaron M, Perret S (2005) Seasonal

timing of rockfall activity on a forested slope at Taschgufer

(Swiss Alps)—a dendrochronological approach. Zeits-

chrift fur Geomorphologie 49:89–106

Venables WN, Ripley BD (2002) Modern applied statistics with

S. Springer, New York

Walker LR, Zarin D, Fetcher N, Myster R, Johnson A (1996)

Ecosystem development and plant succession on landslides

in the Caribbean. Biotropica 28:566–576

Wilson V. 1966. Plant colonization on part of the Hope land-

slide. Master thesis, Department of Geography, Simon

Fraser University

46 Plant Ecol (2012) 213:35–46

123